本文探讨了深度强化学习中的关键技术和挑战，包括延迟稀疏奖励、长期规划、层级结构、子目标设定、部分可观测性、不完美信息处理、大规模状态与动作空间等问题，并介绍了如Hierarchical Deep Reinforcement Learning、Thompson Sampling、Boltzmann Exploration等方法。
1.Delayed, sparse reward(feedback), Long-term planning
Hierarchical Deep Reinforcement Learning, Sub-goal, SAMDP, optoins, Thompson sampling, Boltzman exploration, Improving Exploration
2.Partial observability, Imperfect-Information
Memory, Nash equilibria, MCTS, self-play, LSTM, active perception, curiosity
3.Large state space, Large action space
Hardware, Distributon, Deeper Neural Network.
扫一扫
737
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
